평범한 필기장

Paper : https://arxiv.org/abs/2210.02747 Flow Matching for Generative ModelingWe introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs basearxiv.org(성민혁 교수님 강의 자료 참고 : https://www.youtube.com/watch?v=B4F..

Paper : https://arxiv.org/abs/2304.08465 MasaCtrl: Tuning-Free Mutual Self-Attention Control for Consistent Image Synthesis and EditingDespite the success in large-scale text-to-image generation and text-conditioned image editing, existing methods still struggle to produce consistent generation and editing results. For example, generation approaches usually fail to synthesize multiple imaarxiv.o..

Paper : https://arxiv.org/abs/2310.01506 Direct Inversion: Boosting Diffusion-based Editing with 3 Lines of CodeText-guided diffusion models have revolutionized image generation and editing, offering exceptional realism and diversity. Specifically, in the context of diffusion-based editing, where a source image is edited according to a target prompt, the process comarxiv.orgProject Page : https:..

Paper : https://arxiv.org/abs/2401.11739 EmerDiff: Emerging Pixel-level Semantic Knowledge in Diffusion ModelsDiffusion models have recently received increasing research attention for their remarkable transfer abilities in semantic segmentation tasks. However, generating fine-grained segmentation masks with diffusion models often requires additional training on anarxiv.org0. Abstract Diffusion m..

Paper : https://arxiv.org/abs/2303.04761 Video-P2P: Video Editing with Cross-attention ControlThis paper presents Video-P2P, a novel framework for real-world video editing with cross-attention control. While attention control has proven effective for image editing with pre-trained image generation models, there are currently no large-scale video gearxiv.orgGithub : https://github.com/dvlab-resea..

0. AbstractKey Challenge : Naive DDIM inversion process의 각 step에서의 randomness와 inaccuracy에 의해 발생하는 error를 제한하는 것.이는 video editing에서 temporal inconsistency를 야기할 수 있다.1. Introduction 본 논문은 diffusion model을 이용해서 zero-shot video editing method를 만드는 것을 목표로 한다. Inversion process는 temporally cohorent initial latents의 sequence를 제공함으로써 video editing 결과에 도움을 준다. 그러나 아래의 이미지처럼 direct inversion process는 pot..

Paper : https://arxiv.org/abs/2403.12002 DreamMotion: Space-Time Self-Similar Score Distillation for Zero-Shot Video EditingText-driven diffusion-based video editing presents a unique challenge not encountered in image editing literature: establishing real-world motion. Unlike existing video editing approaches, here we focus on score distillation sampling to circumvent the stanarxiv.orgProject P..

Paper : https://arxiv.org/abs/2408.06070 ControlNeXt: Powerful and Efficient Control for Image and Video GenerationDiffusion models have demonstrated remarkable and robust abilities in both image and video generation. To achieve greater control over generated results, researchers introduce additional architectures, such as ControlNet, Adapters and ReferenceNet, to intearxiv.orgGithub : https://g..

Paper : https://arxiv.org/abs/2403.07420 DragAnything: Motion Control for Anything using Entity RepresentationWe introduce DragAnything, which utilizes a entity representation to achieve motion control for any object in controllable video generation. Comparison to existing motion control methods, DragAnything offers several advantages. Firstly, trajectory-based isarxiv.orgProject Page : https://..

Paper : https://arxiv.org/abs/2311.17338 MagDiff: Multi-Alignment Diffusion for High-Fidelity Video Generation and EditingThe diffusion model is widely leveraged for either video generation or video editing. As each field has its task-specific problems, it is difficult to merely develop a single diffusion for completing both tasks simultaneously. Video diffusion sorely relyinarxiv.org1. Introduc..